Object-Centric Occupancy Estimation for Accurate 3D Vehicle Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing occupancy estimation models in vehicles struggle to accurately identify objects in three-dimensional space, leading to inefficiencies in autonomous or semi-autonomous vehicle operations.
Innovation Solution
A method involving an occupancy estimation model that combines occupancy loss and object loss to refine parameters, using weighted cross-entropy and shape regression to improve object detection accuracy, and employs an adaptive probability threshold for enhanced object identification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If existing occupancy estimation models are used, then the system can process three-dimensional space data, but the object detection accuracy is insufficient
Solution Approach 1:
The patent combines occupancy loss and object loss into a unified total loss function for joint optimization. The occupancy estimation model and object detection model share parameters and are trained simultaneously, allowing the model to learn both occupancy probabilities and object boundaries together, thereby improving detection accuracy and reliability
Solution Approach 2:
The patent introduces an adaptive probability threshold that dynamically adjusts based on object characteristics and scene context. By changing the decision threshold parameter adaptively rather than using a fixed value, the system improves its ability to accurately identify objects while maintaining reliability across different operating conditions
Data Source
AI summary
A method of updating parameters of an occupancy estimation model. The method includes receiving images that are two-dimensional. An occupancy estimation model is utilized to generate a voxel based on the images. An occupancy loss is determined by comparing the voxel to an occupancy ground truth corresponding to the voxel. An object loss is determined by comparing the voxel to an object ground truth. The object loss is combined with the occupancy loss to determine a total loss for the voxel. Parameters of the occupancy estimation model are updated to reduce the total loss determined.